A Modular Structured Approach to Conditional DecisionTheoretic Planning
Liem Ngo Peter Haddawy Hien Nguyen Decision Systems and Artificial Intelligence Lab Dept. of EE&CS, University of WisconsinMilwaukee
A realistic system for planning with uncertain in formation in partially observable domains must be able to reason about sensing actions and to condition its further actions on the sensed infor mation. Among implemented planning systems, we can distinguish two approaches to contingent decisiontheoretic planning. The first is char acterized by a highly unconstrained plan space, while the second is characterized by a constrained and inflexible specification of plan space. In this paper, we take a middle ground between these two approaches that we consider to be more prac tical. We permit the user to specify the structure of the space of possible plans to be considered but to do so in a flexible manner. This flexibility is obtained through the use of a modular represen tation. We separate the representation of actions from the representation of domain relations and we separate those from the representation of the plan space. Actions and domain relations are represented with schematic Bayes net fragments and plan space is represented using programming lan guage constructs. We present a planning system that can find optimal plans given this represen tation.